[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Octave-bug-tracker] [bug #32438] Octave crashes in matrix (bigger than
From: |
José Luis García Pallero |
Subject: |
[Octave-bug-tracker] [bug #32438] Octave crashes in matrix (bigger than 51x51) multiplication in Debian GNU/kFreeBSD |
Date: |
Sat, 12 Feb 2011 19:34:16 +0000 |
User-agent: |
Mozilla/5.0 (X11; U; GNU/kFreeBSD i686; es-ES; rv:1.9.1.16) Gecko/20110107 Iceweasel/3.5.16 (like Firefox/3.5.16) |
Follow-up Comment #2, bug #32438 (project octave):
John, you are right. The problem is ATLAS. I can't check the 3.4.0 version
because I have no time to recompile it with a reference BLAS, but if I install
the Octave 3.2.4 from the Debian kFreeBSD repos and I use only the reference
BLAS the error dissapears.
In Debian GNU/kFreeBSD repositories is provided the 3.8.3 version of ATLAS
that is the same as in the Debian GNU/Linux repos. But I don't know the reason
for the correct behavior in Linux and incorrect in kFreeBSD. I'm not a user of
pure FreeBSD system so I don't know if the error is present there. I'll send a
message in the ATLAS mailing lists.
Here is a small example in C to check the behavior of ATLAS (modify the
#define N line in order to see the results: with 51 runs but with 52 or more
crash in Debian GNU/kFreeBSD):
#include<cblas.h>
#define N 52
int main()
{
double A[N*N],B[N*N],C[N*N];
cblas_dgemm(CblasColMajor,CblasNoTrans,CblasNoTrans,N,N,N,1.0,A,N,B,N,1.0,C,N);
return 0;
}
Compiling order: gcc test.c -o test -lcblas
_______________________________________________________
Reply to this item at:
<http://savannah.gnu.org/bugs/?32438>
_______________________________________________
Mensaje enviado vía/por Savannah
http://savannah.gnu.org/